From Loops to Oops: Fallback Behaviors of Language Models Under Uncertainty
This work addresses the problem of undesirable model behaviors for AI safety and reliability, though it is incremental in analyzing existing phenomena.
The paper investigates how large language models (LLMs) exhibit fallback behaviors like sequence repetitions, degenerate text, and hallucinations under uncertainty, finding a consistent ordering where more advanced models shift from repetitions to hallucinations, and decoding techniques can trade one unwanted behavior for another.
Large language models (LLMs) often exhibit undesirable behaviors, such as hallucinations and sequence repetitions. We propose to view these behaviors as fallbacks that models exhibit under epistemic uncertainty, and investigate the connection between them. We categorize fallback behaviors - sequence repetitions, degenerate text, and hallucinations - and extensively analyze them in models from the same family that differ by the amount of pretraining tokens, parameter count, or the inclusion of instruction-following training. Our experiments reveal a clear and consistent ordering of fallback behaviors, across all these axes: the more advanced an LLM is (i.e., trained on more tokens, has more parameters, or instruction-tuned), its fallback behavior shifts from sequence repetitions, to degenerate text, and then to hallucinations. Moreover, the same ordering is observed during the generation of a single sequence, even for the best-performing models; as uncertainty increases, models shift from generating hallucinations to producing degenerate text and finally sequence repetitions. Lastly, we demonstrate that while common decoding techniques, such as random sampling, alleviate unwanted behaviors like sequence repetitions, they increase harder-to-detect hallucinations.